management, control nodes, portfolio managers, etc.). This information should reveal several broad themes. In particular, it should allow the user to be conclusive concerning: II Whether the manager is generating a forecasted level of tracking error that is consistent with the target established by the mandate. II Whether, for each portfolio taken, individually and for the sum of all portfolios taken as a whole, risk capital is spent in the expected themes. II Whether the risk forecasting model is behaving as predicted. Is the Forecasted Tracking Error Consistent with the Target? The forecasted tracking error is an estimate of the potential risk that can be inferred from the positions held by the portfolio derived from statistical or other forward-looking estimation techniques. An effective risk process requires that portfolio managers take an appropriate level of risk (i.e., neither too high nor too low) vis a vis client expectations. This forecast should be run for each individual portfolio as well as for the sum of all portfolios owned by the client. Tracking error forecasts should be compared to tracking error budgets12 for reasonableness. Policy standards should determine what magnitude of variance from target should be deemed so unusual as to prompt a question and what magnitude is so material as to prompt immediate corrective action. In this manner, unusual deviations across accounts will be easier to identify. Figure 17.1 is an example of a tracking error forecast report for a sample U.S. equity fund produced by Goldman Sachs Asset Management (GSAM) on its proprietary portfolio analysis and construction environment (PACE) platform. PACE is a risk and return attribution system that we use to forecast risk across the spectrum of equities managed by GSAM. Observe from the header of this report that the forecasted tracking error for this account, as estimated by the PACE model, is 3.68 percent per annum. A second equity factor risk model, Barra, projects a tracking error forecast of 2.57 percent. Since each model uses different assumptions to forecast risk, it is not surprising that two different models would produce different results. What is comforting in this case is that both measures of risk are comparable to the targeted risk level of 3.25 percent per annum. This same report should be produced for each account that is supposed to be managed in a parallel manner to ensure consistency of overall risk levels. Is Risk Capital Spent in the Expected Themes for Each Portfolio? In financial variance monitoring, it is insufficient to know only that the overall expense levels are in line with expectations. Each line item that makes up the total must also correspond to expectations. If there are material variances among line items that tend to offset each other, the person monitoring variances should be on notice that unusual activity may be present. As an example, if a department meets its overall ex- 12Tracking error budgets should exist for each portfolio and be determined as part of the organization's asset allocation process.